Raw Dynamic Range

By Rishi Sanyal

The Raw dynamic range capabilities of the Canon 5DS and 5DS R are on the minds of perhaps many of the cameras' target users. While studio photographers who shape their own light may be somewhat ambivalent over the entire topic, landscape, wedding, event, and generally speaking 'available light' photographers are often interested in the ability to push and pull shadow detail in Raw files, or simply 'rescue' underexposed shots. We'll take a bit of a deep dive into the dynamic range capabilities of Canon's new megapixel monsters below, and will be using the 5DS R for all of our tests (the 5DS should have identical results).

Note: Comparisons are slightly complicated by the aggressive tone curve ACR is applying to the 5DS R files that are crushing its blacks, and potentially decreasing the levels of visible noise by making them darker. We expect a later version of ACR to fix this, at which point we will re-process and revisit these results.

Exposure Latitude

In our first test, we look to see how tolerant of pushing exposure the 5DS R's Raw files are. We've done this by exposing our scene with increasingly lower exposures at the camera's base ISO of 100, then pushed them back to the correct brightness using Adobe Camera Raw. Examining what happens in the shadows allows you to assess the exposure latitude (essentially the dynamic range) of the Raw files.

There are two things at play here: less exposure means less light is captured, which itself leads to noisier images because of shot noise. Hence, larger pushes of progressively underexposed files will always be noisier than properly exposed shots. However, a camera's sensor performance can also affect results by adding additional noise (or not), which leads to the difference in performance you'll see between cameras with the same sensor size in our widget.

The 5DS R appears to hold up as as well as the 6D with respect to exposure latitude, when normalized to common viewing size. This is definitely a step in the right direction for Canon, as the 6D already showed some improvements over the 5D Mark III when it came to dynamic range. Furthermore, it's nice to see that the increased pixel count, which might have led to an increase in total read noise due to the extra pixels that have to be read, does not appear to adversely affect dynamic range.

The burning question here though, for some, is how the 5DS cameras hold up to the competition. In a nutshell, these cameras can't compete with the impressive dynamic range, and therefore exposure latitude, of cameras like the Nikon D750, which is tolerant of even 4 to 6 EV pushes of base ISO shots that tend to show unacceptable levels of noise with the 5DS R. That said, the 5DS R does hold up well enough to more smaller pushes of 2 to 3 EV, although a close look at the pixel-level will show noise creeping into shadows that wouldn't be there in a camera with better sensor performance (lower read noise).

ISO Invariance

Since the exposure latitude test above conflates both shot noise and read noise performance into an assessment of how recoverable a camera's shadows are at base ISO, we further investigate the camera's sensor and electronic performance by holding focal plane exposure the same, and assessing the benefit of in-camera ISO amplification to overcome camera read noise. A camera with very low read noise (a better performer) will show little to no benefit to performing this amplification, or image brightening, in-camera vs. in post-processing, and will therefore be 'ISO-invariant' (or 'ISO-less'). A camera with higher read noise (a worse performer) will show significant benefit to performing this amplification in-camera as opposed to in post-processing, and will therefore be 'ISO-variant'.

In other words, a camera with high base ISO dynamic range that comes from having a low noise floor can reduce the need to amplify the sensor's signal to keep it above that noise floor even under conditions traditionally demanding higher ISOs.

The benefit? Increased highlight retention ability if you shoot Raw and purposefully keep your ISO setting low so that you can brighten shadows and midtones in your favorite Raw converter, while holding back highlights and keeping them from clipping.

Below we've used the same aperture and shutter speed at different ISO settings to see how much difference there is between shooting at a particular ISO setting (and using hardware amplification) or digitally correcting the brightness, later.

Above, you'll want to compare each pushed shot to the ISO 6400 shot at the bottom right, which is the reference, or 'best case'. An ISO-invariant camera will show little to no difference across all crops, while an ISO-variant camera will.

The Canon 5DS R, like its predecessors, is ISO-variant, as pushed lower ISO files are noisier than their higher ISO counterparts, when the camera is given exactly the same exposure. The ISO 100 and 200 shots pushed by 6 and 5 EV in post, as opposed to in-camera by setting the ISO to 6400, show much higher noise levels across almost all tones. ISO 400 pushed by 4 EV still shows worse performance than the same exposure pushed in-camera via ISO amplification. By ISO 800 the differences become marginal, although you may see a slight increase in shadow noise with the 3 EV push in-post as opposed to in-camera. This performance is an improvement over the 5D Mark III and perhaps even over the 6D, both of which still have noisier midtones and shadows at ISO 800 with a 3 EV push compared to ISO 6400. Comparing the 5DS R to its predecessors side-by-side at common viewing size, it's quite clear that Canon is making progress, with significant improvement over the 5D Mark III. Note the lack of banding in the 5DS R compared to the 5D Mark III, which really helps with shadow recoverability. The 5DS R appears to even slightly outperform the 6D, which is impressive given its resolution.

Compared to its immediate competition, though, the 5DS R is not anywhere near as ISO-invariant as a Nikon D810 or D750. Nor is it as ISO-invariant as the Sony a7S, which is itself one of the underperformers in this regard from Sony (the a7R performs better). The latest offerings from Nikon and other camera companies sporting Sony imaging sensors are becoming ISO-invariant enough that there's very little benefit to performing image brightening in-camera by raising the ISO setting (holding shutter speed and aperture constant), which means you can save yourself stops of highlight information under situations traditionally demanding higher ISO settings by keeping the ISO low and brightening everything but the highlights in post-processing of the Raw file. You can think of this as a better way to brighten your image if you're worried about clipped bright tones, by aggressively rolling off highlights - something that simple brightening via ISO ampliication in-camera cannot do. An ISO-invariant camera allows you to do this without the noise penalty you see in the lower ISO shots of the 5DS R, above.

This test essentially tells us a lot about the sensor and camera's electronics. Using the same exposure means that all the images were created from the same amount of total light (so have the same shot noise). This means that any differences in noise must be the result of read noise added by the camera. Ultimately, the 5DS cameras continue to show higher levels of read noise compared to their peers, likely due to continued use of analog-to-digital converters (ADC) on chips external to the image sensor itself. The extra read noise means that you won't be able to use less hardware amplification to retain extra highlight information as well as you could with some of the cameras' peers, and you'll have to be more careful about ensuring proper exposure and ISO at the time of capture.